This Big Technology Podcast episode interviews OpenAI's Chief Research Officer, Mark Chen, about the release of GPT 4.5. Chen explains GPT 4.5 as the latest step in OpenAI's predictable scaling paradigm, representing a significant improvement over previous models. He discusses the parallel development of reasoning models alongside scaling efforts and refutes the idea of a "scaling wall," asserting continued progress in model capabilities. Chen highlights GPT 4.5's superior performance in everyday use cases (60% preference rate) and productivity tasks (70% preference rate) compared to GPT 4.0, particularly in creative writing and certain scientific domains. The interview also touches upon model efficiency, talent retention at OpenAI, and the future of AI, emphasizing the importance of both large, general-purpose models and smaller, niche models.
